Infrasound from Tornados: Theory, Measurement, and Prospects for Their Use in Early Warning Systems

Carrick Talmadge Tornados may produce a low-frequency signature that could be used in automatic warning systems. Postal: National Center for Physical Acoustics Introduction University of Mississippi Infrasound is propagating sound waves with frequencies below the range of human University, MS 38677 hearing. A practical frequency range for infrasound that will propagate over long USA distances is 0.01-20 Hz. The lower frequency cut off for propagating infrasound Email: arises because the buoyancy of parcels of air becomes comparable to the pressure [email protected] gradient forces of acoustic waves at sufficiently low frequencies. The precise -fre quency for this cut off occurs at the Brunt–Väisälä frequency (Stull, 1995), the Roger Waxler exact value of which depends on details of the vertical profile of the atmosphere. Postal: One of the more famous examples of long-range propagating infrasound is the ex- National Center for Physical Acoustics plosive eruption of Krakatoa Volcano in Indonesia. Audible sounds were heard as University of Mississippi far away as 5,000 km, and low-frequency infrasound signals with periods greater University, MS 38677 than one minute were propagated around the Earth at least seven full times (e.g., USA Fee and Matoza, 2013). Email: [email protected]

Figure 1. Annual average death rate from tornados. Also shown are the geographical regions referred to in the text. Note that these geographical regions correspond approximately to tornado climatological zones rather than to traditional geographical regions. Adapted from original figure by The National Oceanic and Atmospheric Administration (NOAA).

Because infrasound propagates for long ranges, it can be used for hazard monitoring. For volcanic warnings, atmospheric blasts can be detected by infrasound stations even when the peak of the volcano is obscured by clouds so that optical observations are not possible. Another important infrasound source is hurricanes, which produce characteristic tonelike signals called “microbaroms” produced by

the nonlinear interaction between ocean waves and the atmosphere (Waxler and Gilbert, 2006). It was reported by Raveloson et al. (2012) that infrasound signals from the 2011 Tohoku-Oki, Japan, earthquake could have been used as an early warning of the impending tsunami. Tornados represent one of the most common natural hazards posed in the United States. Within any given year, on average, some 800 tornados will occur within the United States east of the Rockies, resulting in 80 deaths and 1,500 injuries [National Oceanic and Atmospheric Administra- tion (NOAA) National Severe Storms Laboratory]. In spite of the mystique associated with “Tornado Alley” (typically listed as the states of Oklahoma, Kansas, and Nebraska as Figure 2. Radar reflectivity for the Texas supercell over Wichita Falls, well as adjoining areas from neighboring states), southern TX on April 11,1979. The location of the mesocyclone is shown by the states, including Mississippi, remain a primary target for circle, which is oriented over the hook region of the storm. Obtained tornadic activity. These regions are shown inFigure 1. from NOAA, Don Burgess, OSF, and Vanessa Ezekowitz. Even with recent advances in Doppler radar technology and other early warning systems, tornados in the central regions our recent tornadic thunderstorm measurements in Okla- of the United States remain a significant risk for injury or homa (Frazier et al., 2014). Finally, we show that there are loss of life. Augmenting existing detection arrays with infra- characteristic signals that seem to be present only when tor- sound/low-frequency arrays offers a possible new method nados have touched down. It is possible that these character- for improving the safety of people living in high-risk tor- istic signals could be used to augment existing early warning nado areas. systems and lead to an improvement in safety for people at risk from tornadic thunderstorms. We focus here on the possible generation of infrasound from tornados and detection of these waves on a regional scale Tornado Genesis (i.e., over distances up to 100 km from the source). For this There are two generally accepted patterns for energetic tor- distance scale, the main influences on infrasound propaga- nado generation (reviewed in Bluestein, 1999). The first is tion are the distance of the source and the effective vertical associated with supercell thunderstorms (storms that are atmospheric sound-speed profile in the direction of prop- large, typically 15 km wide by 15 km tall) that have a deep, agation (e.g., Attenborough et al., 2006). In a (theoretical) continuously rotating updraft known as the mesocyclone. atmosphere with a constant vertical sound speed, sound Tornados from supercells are the most studied class of tor- intensity decreases as the inverse square of the propagation nados due to the relative predictability of where a tornado distance. For “downwind” propagation—known as “ducted” will appear within the supercell (the “hook region” in Figure propagation—sound intensity decreases inversely with dis- 2). These are also studied in part because of the visibility of tance. Hence it is important for long-distance propagation supercells, which often have localized precipitation zones measurements to place the arrays downwind from the infra- separated from the mesocyclone (a large region of rotation sound source when possible. within a thunderstorm) and because of the geographical ac- cessibility of the Great Plains region where many of this type In this paper, we start by discussing the mechanisms for tor- of storm occur. Supercells are noted for producing the most nado genesis. We then review the hazards associated with intense and dangerous tornados, with nearly half of all fa- tornados, noting that while the Great Plains (a grassland talities occurring from these intense storms. However, even prairie east of the Rocky Mountains that extends from the with their relatively well-defined structure, current radar al- southern US border into Canada) have received much of the gorithms achieve less than a 50% probability of detection for attention for tornadic research, there is a substantial safety acceptable levels of a false alarm rate (Mitchell, 1998). risk from tornados in the “US Mid-South” (Figure 1). We then discuss evidence that tornado vortices produce infra- The other mechanism for tornado genesis is simply referred sound and low-frequency sound. We review the results from to as “non-supercell.” Typically, these tornados are associ-

44 | Acoustics Today | Spring 2016 ated with less organized storm systems, in which precipitation bands shroud the region of formation of the tornado. Tornados generated by this mechanism tend to be weaker, but they are very dangerous due to their lack of predictability and the poor visibility associated with the phenomenon. Roughly one half of the storms in the Southeastern United States are of this variety, with almost 100% of the tornados spawned in Florida being non-supercell in origin (Kelly et al., 1978).

Tornado Alley Tornados in the US Plains States represent an all too famil- iar threat to life, health, and property. What is not typically well appreciated is that the area that has the largest threat for loss of life is the United States Mid-South. Contributors Figure 3. Histogram of tornadic events for (a) the "Mid-South" geo- to the higher mortality rate include rain-wrapped thunder- graphical region and (b) the “South-Central” geographical region. storms, which obscure tornados within the storm cell as well The states used for each region are shown by their state abbreviations. more convoluted terrain and the presence of tall forests that The green shaded area is for all tornadoes, and the magenta area is obscure the horizon. Without advanced warning from the for just violent (EF3-EF5) tornados. emergency broadcast system that provides alerts to oncom- ing tornados, people who live in rural areas in these states are subject to the whims of nature. In spite of the South-Central region’s deserved reputation In the Great Plains, low-precipitation supercells are a rela- for having a high rate of tornados, we see that the Mid-South tively common occurrence (Bluestein, 1999). Because of the has nearly identical rates for tornados as the South-Central, lack of significant precipitation, very clear views of tornadic and both the East-North-Central (Figure 1) and Mid-South activity are usually possible, making them ideal for the study regions suffer higher fatality rates than does the South-Cen- of tornados. The relatively smooth terrain and lack of tall tral region. Lower visibility due to terrain, presence of trees, forests also allow views to the horizon in many cases. and a higher frequency of rain-wrapped tornados are likely contributors to these higher fatality rates. Possibly the fact In Figure 3 we show histograms of the total number of tor- that Plains States tornados and the meteorological factors nados for each day over the period 1950–2009 summed over associated with their genesis are well-studied has reduced the South-Central and over the Mid-South regions (as de- the risk of fatality associated with South-Central tornados. fined in Figure 1). These figures show that the tornado season is roughly concentrated in the band from days 80–180 Infrasound from Tornados of the year (mid-March through the end of June of each Since the early 1970s, convective systems have been known year). For the Mid-South, there is strong activity from Janu- to be sources of infrasound (e.g., Bedard and Georges, 2000). ary through March but also late season activity in November The typical frequency band for these sounds is about 0.017 and December. –1 Hz (Georges, 1973). However, there is a great deal of un- The relatively narrow window for tornadic activity and a certainty about the origins and significance of the sounds. A safer operating environment, due to good visibility and road variety of potential source mechanisms have been proposed systems that are on a regular grid, make the South-Central for infrasound generated by convective storms. Georges region the preferred region for tornado research and tor- (1976) considered many of these and concluded that vortex nado chasing. The downside of this circumstance is that motion was the most plausible candidate. While the sug- most research on tornado genesis arises from data obtained gestion has been made that these signals are a precursor to from low-precipitation supercells, and tornado genesis from tornadic activity, Bedard (2005) concluded that “it seems high-precipitation supercells and front boundaries (more unlikely that the much lower frequencies detected by these common in the Mid-South) tend to be understudied (which geoacoustic observatories had any direct connection with ultimately may contribute to the higher mortality rates re- tornado formation.” corded in the Mid-South). Spring 2016 | Acoustics Today | 45 Infrasound from Tornadoes

The first published accounts of sound recordings from a tor- Figure 4. Digital NCPA sensor nado were by Arnold et al. (1976). These data were restricted (for scale, floor tiles 1 foot × 1 foot to the frequency range from 100–2,000 Hz. All measure- in size). The sensor is GPS syn- ments were collected using low-quality analog equipment, chronized, 1000 samples/second, (22.5 noise free bits). which led to significant distortion of the measured signal. The main findings were that the measured signal was broad- ure 4) developed at the NCPA. band in character, with a roll-off above 500-Hz, and the band These sensors have a built-in start frequency was associated with the lower frequency cut 24-bit digitizer with GPS time off of the measuring equipment. syncing and 802.11/b wireless More recently Bedard (2005) has reported on measurements connectivity. of infrasound by a four-element array of sensors in the 0.5– The NCPA sensors have a unique capability to measure sig- 10 Hz band from tornadic thunderstorms and concluded nals from infrasound to low-frequency acoustics (0.015-500 that the 0.5–2.5 Hz band contained the maximum correla- Hz). The frequency response of this sensor was from 0.001- tion between sensors. Note that this result is dependent on 100 Hz (Figure 5), and its gain was set so that the maximum the sensor separation, the atmospheric noise field at the time transducible pressure amplitude was about 90 Pa. For this of the measurements, and characteristics of the sensors (such exercise, the sensor sampling rate was set to 1,000 samples/ as its frequency response and noise floor), as well as on the second. characteristics of the infrasound signal at the array. It was suggested that this signal was associated with the tornadic activity, but the mechanisms for its creation remain unclear. For example, Abdullah (1966) proposed a mechanism based on radial oscillations of the tornado vortex; however, Shecter (2012) demonstrated that this mechanism was implausible. Other possible mechanisms include turbulent flow associated with the tornado vortex and interactions between the tornado and the ground. If tornados were known to produce infrasound or low-frequency signals, and if these signals had characteristic fea- tures, infrasound arrays could be used to augment existing Figure 5. Nominal sensitivity curve for the NCPA microphones used early warning systems and possibly significantly improve in this deployment. This sensitivity function has an asymptotic value the safety of individuals living in high-risk areas for torna- of 25 mV/Pa, a pole at 6.4 mHz, and a zero at the origin. dos. However, previous results do not unequivocally establish that tornados produce infrasound nor do they provide Oklahoma Campaign any characteristics of the signal that are unique to tornados Oklahoma was targeted for this project because the storm- that have touched down. As part of a program on hazards chasing season lasts for a shorter period. As mentioned pre- monitoring that were funded by NOAA, the National Center viously, the relative flat terrain coupled with the absence of for Physical Acoustics (NCPA) and Hyperion Technology standing forests offers improved visibility and a reduced risk Group Inc. conducted a series of infrasound sensor deploy- of unintentional tornado intercepts. One of the objectives ments in Oklahoma over the summer of 2011. The primary of this deployment was to obtain broadband signals from interest in this deployment was to address both of those is- tornados by placing sensors in the region of tornadogenic sues. storms. In some cases, sensors were placed within a few ki- lometers of the path of very large tornados (EF4 and EF5), Data Collection but this was not a primary goal of the project. The NCPA, in collaboration with Hyperion Technology Group Inc., collected data for two tornado outbreaks on May The second and primary objective was to deploy region- 24, 2011, and June 11, 2011. The sensors used for these mea- al arrays northeast of the tornadic convective storm. The surements were a new class of digital infrasound sensor (Fig- northeast direction was chosen because this is the direction

46 | Acoustics Today | Spring 2016 of travel of the tornados and, in principle, would allow ob- servation of the tornado from its touchdown location and along the path of the tornado as the tornado approached the regional array. We would expect sound ducting to also be dominantly in the northeast direction. In order to simplify the installation process, the array elements should be equally spaced on public easements along roads. The spacing was chosen to be 1 km based on the reported infrasound by Bedard (2005) in the 0.5-2.5 Hz band. This spacing turned out to be too large because no infrasound below a few hertz was observed in this deployment. The sensor deployment consisted of two teams of research- Figure 6. Array locations and main storm tracks. ers from Hyperion Inc.: one that would be responsible for the regional arrays and a second that would aim for near- field intercepts of tornados. The locations of the arrays were Table 1: Major tornados in Oklahoma during the May 24, 2011, outbreak; also included is the Stillwater tornado. Data are from the selected based on severe weather forecasting. Table 1: Major tornados in Oklahoma during the 24 May 2011 outbreak; also included is the NormanStillwater OK tornado. National Data are Weatherfrom the Norman Service OK National Forecast Weather Office.Service Forecast Office. The findings of this study are reported in Frazier et al. Time Duration Length Width Name Scale (2014). Here we summarize the main results, focusing on [GMT] [minutes] [miles] [yards] data and observations that point to the tornado vortex as Canton Lake EF-3 15:20-15:43 23 13 880 the main candidate for the source of the infrasound and low Calumet-El Reno- frequency reported in Frazier et al. (2014). Piedmont-Guthrie EF-5 20:50-22:35 105 63 1760 (CEPG) Chickasha-Blanchard- May 24, 2011, Oklahoma Outbreak EF-4 22:07-23:01 54 33 880 Newcastle (CBN) The May 24, 2011, outbreak, was part of a series of violent weather events that extended from May 11–26. During this Lookeba EF-3 20:31-20:46 15 9 880 period, a total of 244 tornados were recorded in 12 states, of which two were EF-5 (>200 mph estimated winds), three Stillwater (STW) EF-2 22:50-23:05 15 19 880 were EF-4 (166–200 mph estimated winds) and eight were Washington-Goldsby EF-4 22:27-23:05 38 23 880 EF-3 (115–135 mph estimated winds) in strength (https:// www.ncdc.noaa.gov/data-access/severe-weather). This outbreak included the deadly May 22 tornado in Joplin, Mis- A severe EF-5 tornado (denoted as CPEG in Table 1) touched souri. During this outbreak there were 178 fatalities and down south of Hinton, OK, and remained on the ground 1,629 reported injuries associated with tornadic activity and for 63 miles, taking approximately 105 minutes to cover this the economic damage from this outbreak exceeded $7 bil- distance. A speed of 151 mph, the highest wind gust ever lion US (Buhayar, 2011). A Google Earth KML file of the registered on an Oklahoma Mesonet station1, was generated tornado track data is provided at http://www.srh.noaa.gov/ by this tornado as it passed El Reno, OK. Approximately 15 oun/?n=events-20110524. minutes after the CPEG tornado lifted, a new EF-2 tornado The major tornados in Oklahoma during the May 24, 2011, (denoted STW) was spawned from a separate supercell and outbreak are shown in Table 1. Tracks associated with torna- touched down south of Stillwater, OK, almost in line with do intercepts are shown in Figure 6 together with the array the CPEG tornado. This tornado produced a 19-mile track locations where data were collected for these events. As can and remained on the ground for 15 minutes. be seen, this deployment was highly successful in locating arrays in both the near field for the Chickasha-Blanchard- 1 The Oklahoma Mesonet is a network of 120 automated environmen- Newcastle (CBN) tornado, and regionally for the Calumet- tal monitoring stations developed by the University of Oklahoma and El Reno-Piedmont-Guthrie (CPEG) and Stillwater (STW) Oklahoma State University. tornados.

Spring 2016 | Acoustics Today | 47 Infrasound from Tornadoes

Through a combination of luck and good planning, Array 1 (shown in Figure 7) was optimally positioned to measure the infrasound from both tornados. As discussed below, the lifting of the CPEG tornado, followed after an interlude by a second less-intense tornado, provides clear evidence that infrasound and low-frequency sound detected during these tornado events are associated with the tornadic activity. The meteorology and infrasound/low-frequency propagation for Array 1 are discussed in Frazier et al. (2014) and they were very favorable for signal detection over the entire paths of the two tornados. A third EF-4 tornado (denoted here as CBN) touched down just southwest of Chickasaw, OK, traveled a distance of 33 miles, and remained on the ground for 54 minutes. Array 2 was placed northeast of this tornado, and as shown in Figure 8, resulted in a near intercept of the tornado by the array with the closest point of approach of the core path ranging from 700 m to 1,200 m of the infrasound sensors. Unfortunately, the sensitivity of the sensors was too high for such a close-in approach, so a portion of the sensor re- Figure 7. Array geometry for Array 1. The nominal location of this cordings clipped when the tornado was near its point of array was 36.10107°N, 97.26492°W, and is about 19 km west of Stillwater. The actual spacing between elements was closer to 500 m. closest approach. Results from Array 1 for the CPEG and STW Tornados The envelope of the pressure signal from element SN056 of Array 1 is shown in Figure 9. As illustrated, the level of sound increases as the CPEG tornado approaches, and there is a sudden drop in sound when the tornado lifts. After the new STW tornado spawns, sound levels again intensify until immediately after the dissipation of that tornado. Sound files from Array 1 for represen- tative periods during this deployment are provided at http://acousticstoday.org/4162-2/.

Figure 8. Array geometry for Array 2. The nominal location of this array was located at 35.19669°N, 97.65315°W in Blanchard, OK, and is about 35 km northwest of Chickasha, OK. As with Array 1, the actual spacing between elements was closer to 500 m. Figure 9. Signal Envelope for SN056.

48 | Acoustics Today | Spring 2016 Additional information can be gleaned by examining the waveform (Figure 10) over the period of the near approach by the CPEG tornado. What is observed, along with the ex- pected strengthening of the signal, are intermittent noise bursts that are clearly coherent over the array. In addition, there are numerous smaller spikes during the period of closest approach. These are not coherent over the array, but a closer inspection of these spikes reveals a waveform resem- bling a damped sinusoidal oscillation with a period of about 200 to 300 Hz. Since this corresponds to the Helmholtz reso- nance of the sensor manifold, the likely explanation is that raindrops striking the top of the sensor caused it to ring at this characteristic frequency.

Figure 11. Position of the CPEG and STW tornados at near approach as well as beam-forming rays (light blue). The time values by each beam-forming ray are the center times at which the beam direction was computed.

tions would yield a net bias in the direction of propagation relative to the true direct of the tornado. This is because the circulation around the tornado will produce a bending of the propagation of the signal as it travels from the tornado vortex to the sensor. A similar effect has also been reported for large tropical cyclones (e.g., Blom, et al. 2014).

Results from Array 2 for the Figure 10. Waterfall plot of signals from all elements of Array 1. The distance from the tornado increases upward. CBN Tornado As described previously, the CBN tornado path was very close to Array 2. As shown in Figure 16 of Frazier et al. The most probable direction of arrival of the infrasound (2014), the beam directions were in good agreement with was estimated using a spatial filtering method called beam the estimated tornado positions. forming (e.g., Krim and Viberg, 1996). This processing was performed over the frequency region 0.5–3 Hz, and it relies Spectral Analysis of the on the infrasound remaining coherent over the infrasound array. Infrasound Signals Coherent infrasound was not observed in the 0.5–2 Hz band The temporal patterns and array beam-forming results pro- reported by Bedard (2005). As pointed out previously, the vide strong evidence that the origin of the infrasound and array geometry was different for this array, which may ac- low frequency observed in the NCPA arrays were associated count for he discrepancy between this and Bedard’s previous with the tornadic vortex. However, spectral analysis can work. yield insights into the mechanisms that generate this infrasound/low-frequency sound. The beam-forming results for CPEG and STW are shown in Figure 11. We found the beam direction was within 10° In Figure 12, we show the spectra as a function of time as of the direction of the estimated position of the tornado, the CPEG tornado approaches Array 1. The predominant which was consistent with our measurement uncertainty of feature observed is the rising high-frequency (80–100 Hz) the beam direction. Assuming the mechanism for produc- shoulder as the tornado approaches. The lifting of the high- ing the observed signal arrives from the tornado vortex, it is frequency tail above 150 Hz is probably associated with the likely that more accurate tornado positions and beam direc- onset of precipitation at the infrasound array. This feature

Spring 2016 | Acoustics Today | 49 Infrasound from Tornadoes

is seen across all results. For the STW and CBN tornados, peak and a ga product that models the effects of attenuation the shoulder diminishes as the tornado moves away from and the propagation losses in the atmosphere (Bass et al. the arrays. Furthermore, during the interlude between the 1995). Example results of this fit are shown in Figure 13. CPEG and STW tornados, this high-frequency shoulder is As discussed further in Frazier et al. (2014), we were con- conspicuously absent (the high-frequency tail thought to be sistently able to accurately model all of the available spectra associated with precipitation is still present however). data using the model in Equation 1.

Figure 12. Spectra as a function of time as the CPEG tornado ap- Figure 13. Measured spectra and best fits using the model from proaches Array 1. Equation 1.

This pattern of a larger high-frequency shoulder when the Summary tornados are nearby is consistent with the mechanism of We have provided evidence that tornado vortices produce generation being associated with the tornadic vortex. As a characteristic spectral signature. Analogous to the “hook described in Frazier et al. (2014), a number of candidate region” used by Doppler radar systems, it is possible that this mechanisms were considered. The only mechanism for this characteristic spectral signature could be used in conjunc- shoulderEquation that we 1 found from to be consistent Talmadge with article: the observa- tion with permanent infrasound arrays to augment existing tions was sound generated by vortex emissions. early warning systems. Clearly, more research would need to be done to establish the viability of such a system. Nonethe- The model we used was based on a jet turbulence model by less, we believe these research results offer a promising new Powel (1959) with terms added to account for the bandwidth direction for further improvements in public safety. nature of the observed turbulence as well as losses associated Equation 1 from Talmadge article: with absorption and propagation of the sound Acknowledgments

ω2ω2 (1) ω = ω−7/3 + As n ω + We would like to acknowledge the employees at Hyperion S ( ,r) Af 4 2 2 2 4 g a ( , r ) A n (1) ω + 4ζ ωnω +ωn Technology Group, LLC for their substantial contributions to this work. This work was funded by the National Oceano- In more detail, Equation 1 models the power spectrum as graphic and Atmospheric Administration. the sum of a ω–7/3 power term associated with Kolmogorov

€ inertial-subrange turbulence (e.g., Shields, 2005). The sec- 2 2 (1) ond term is the −jet7/ 3turbulenceAs ωmodel,nω which has been modi- S (ω,r) = Af ω + ga (ω,r)+ An ω4 + ζ 2ω2ω2 +ω4 fied to include a parameter4 thatn controlsn the width of the

€ 50 | Acoustics Today | Spring 2016

New symbol that appears in following paragraph for your reference:

enlarged version ζ New symbol that appears in following paragraph for your reference:

ζ enlarged version

Biosketches Bedard, Jr., A. J., and Georges, T. M. (2000). Atmospheric infrasound. Phys- ics Today 53, 32–37. Blom, P., Waxler, R., Frazier, W. G., and Talmadge C. (2014). Observations Carrick Talmadge was born in Richmond, of the refraction of microbaroms generated by large maritime storms by IN. He received a BS in Physics from Saint the wind field of the generating storm.Journal of Geophysical Research: Vincent College in 1981 and a PhD at Atmospheres 119, 7179–7192. Purdue University in 1987, focusing on Bluestein, H. (1999). Tornado Alley: Monster Storms of the Great Plains, Ox- ford University Press, New York. experimental gravitational physics. His Buhayar, N. (2011). Joplin Tornado Leads Storms That May Cost Insurers thesis advisor was Ephraim Fischbach. He $7 Billion in One Week. Bloomberg, June, 2011. Available at http://www. remained on staff at Purdue before transferring to the Uni- bloomberg.com/news/articles/201106-06/joplin-tornado-leads-storms- that-may-cost-insurers-7-billion-in-one-week. versity of Mississippi in 1998. He has worked on a variety Fee, D., and Matoza, R. S. (2013). An overview of volcano infrasound: From of areas in science, including experimental gravity research, Hawaiian to plinian, local to global. Journal of Volcanology and Geother- astrophysics, hearing science, low-frequency sound, infra- mal Research 249, 123–139. sound, and metrological sciences (including calibration and Frazier, W. G., Talmadge, C., Park, J., Waxler, R., and Assink, J. (2014). Acoustic detection, tracking, and characterization of three tornadoes. sensor design). Journal of the Acoustical Society of America 135, 1742-1751. Georges, T. M. (1973). Infrasound from convective storms: Examining the Roger Waxler was born and raised in evidence. Reviews of Geophysics 11, 571–594. New York City. He received a BA in Math- Georges, T. M. (1976). Infrasound from convective storms. Part II: A critique of source candidates. NOAA Technical Report ERL 380-WPL 49, NOAA ematics from the University of Chicago Environmental Research Laboratories, Boulder Colorado. Available at and a PhD in Physics from Columbia http://babel.hathitrust.org/cgi/imgsrv/download/pdf?id=psia.ark%3A%2 University, working on theoretical solid F13960%2Ft6nz9gz8m;orient=0;size=100. Kelly, D. L., Schaefer, J. T., McNulty, R. P., and Doswell III, C. A. (1978). state physics with J. M. Luttinger. He be- An augmented tornado climatology. Monthly Weather Review 106, 1172– gan research on acoustics while visiting 1183. Penn State University and then accepted a position in the Krim, H., and Viberg, M. (1996). Two decades of array signal processing National Center for Physical Acoustics (NCPA) at the Uni- research: The parametric approach.IEEE Signal Processing Magazine 13, 67–94. versity of Mississippi. At the NCPA he has been specializing Mitchell, E. D. (1998). The National Severe Storms Laboratory tornado de- in acoustic phenomena and propagation in the atmosphere. tection algorithm. Weather and Forecasting 13, 352–366. Powell, A. (1959). Similarity and turbulent jet noise. Journal of the Acousti- cal Society of America 31, 812–813. References Raveloson, A., Kind, R., Yuan, X., and Ceranna, L. (2012). Locating the Tohoku-Oki 2011 tsunami source using acoustic-gravity waves. Journal of Abdullah, A. J., (1966). The ‘musical’ sound emitted by a tornado.Monthly Seismology 16, 215–219. Weather Review 94, 213–220. Schecter, D. A. (2012). In search of discernible infrasound emitted by nu- Arnold, R. T., Bass, H. E., and Bolen, L. N. (1976). Acoustic spectral analysis merically simulated tornadoes. Dynamics of Atmospheres and Oceans 57, of three tornadoes. Journal of the Acoustical Society of America 60, 584– 27–44. 593. Shields, F. D. (2005). Low-frequency wind noise correlation in microphone Bass, H. E., Sutherland, L. C., Zuckerwar, A. J., Blackstock, D. T., and Hester, arrays. Journal of the Acoustical Society of America 117, 3489–3496. D. M. (1995). Atmospheric absorption of sound: Further developments. Stull, R. B. (1995). Meteorology Today for Scientists and Engineers. Brooks Journal of the Acoustical Society of America 97, 680–683. Cole, Belmont, CA. Bedard, Jr., A. J. (2005). Low-frequency atmospheric acoustic energy associ- Waxler, R., and Gilbert, K. (2006). The radiation of atmospheric micro- ated with vortices produced by thunderstorms. Monthly Weather Review baroms by ocean waves. Journal of the Acoustical Society of America 119, 133, 241–263. 2651–2664.

Spring 2016 | Acoustics Today | 51